Skip to content

Latest commit

 

History

History
65 lines (52 loc) · 1.64 KB

INSTALL.md

File metadata and controls

65 lines (52 loc) · 1.64 KB
Raw

Installation

Prerequisites

All the codes are tested in the environment:

  • Linux (tested on Ubuntu 16.04/18.04/20.04)
  • Python 3.8 (waymo-open-dataset 2-5-0 required)
  • PyTorch 1.10 or higher
  • CUDA 11.0 or higher
  • GCC 5.4+
  • spconv v2.x

Recomended Steps

a. Create a conda virtual environment.

  conda create --name detzero python=3.8
  conda activate detzero

b. Install PyTorch and torchvision following the official instructions.

  pip install torch==1.10.0+cu111 torchvision==0.11.0+cu111 torchaudio==0.10.0 -f https://download.pytorch.org/whl/torch_stable.html

Note: Make sure that your compilation CUDA version and runtime CUDA version match.

c. Install cmake.

  conda install cmake

d. Install sparse conv.

  pip install spconv-cu111

e. Install pytorch scatter (for DynamicVFE). We suggest to follow the instructions of torch_scatter to install the package based on your own environment version.

f. Install Waymo evaluation module.

  pip install waymo-open-dataset-tf-2-5-0

g. Install other required dependent libraries.

  cd DetZero && pip install -r requirements.txt

h. Compile other libraries.

  cd DetZero/utils && python setup.py develop

i. Compile the libraries of specific algorithm modules.

  cd DetZero/detection && python setup.py develop
  cd DetZero/tracking && python setup.py develop
  cd DetZero/refining && python setup.py develop